Why do you have just two numbers to represent an image in the end? shouldn't there be more? Also wouldn't it be a good idea to use autoencoders to create another model to recreate the initial image from the last representative outputs..? So that we could improve the model into one that yields a more representative final representation array.
![“อ.เบียร์ คนตื่นธรรม” มาตามคำเรียกร้อง แหกศาสตร์ฮวงจุ้ย เตือนสติอย่างมงาย | แฉ 7 พ.ย. 67 [1/3]](http://i.ytimg.com/vi/5fc4S3xBNfc/mqdefault.jpg)
Great tutorial and step by step explanation for Siamese Networks. Keep up the good work!
are you a god? For me, your video came like one. Thank you
Great work. Thanks for the tutorial.
Glad it was helpful!
Great video. Thanks for the video. How do you evaluate the performance of the model?
really good explanation, really appreciate
Great tutorial , how to calculate similarité in place dissimilarity?
I got the idea.
But how does backpropagation works if Contrastive Loss is not a network, just some code?
And what is derivative of abs(x) ?
This is really great. Please could I get a tensorflow implementation?
Why do you have just two numbers to represent an image in the end? shouldn't there be more? Also wouldn't it be a good idea to use autoencoders to create another model to recreate the initial image from the last representative outputs..? So that we could improve the model into one that yields a more representative final representation array.
Great thoughts, however this dude just giving a tutorial. Dont be so harsh, you can try those as well.
U make comparison between these two vectors
Thank you, where can i fiund the dataset ?
how to calculate final accuracy
Code is not available?
Your audio is wayyyyyyy too quiet. I had to basically set my volume to max